The AI Journal

In 2006, a radiocarbon lab in Denmark put a date on a sixteen-centimetre twig and moved the chronology of the entire Aegean Bronze Age. The twig was an olive branch, excavated still rooted in its growing position on Santorini, buried alive by the eruption that destroyed the Bronze Age town at Akrotiri. Walter Friedrich and five colleagues wiggle-matched a sequence of radiocarbon measurements along its rings against the international calibration curve and announced the eruption at 1627 to 1600 BC, with 95.4 percent probability, about a century earlier than the date archaeologists had been using on the strength of Egyptian-linked pottery synchronisms (Friedrich et al. 2006). The confidence attached to the new date was unusual: not a bracket of centuries, which is normal for this period, but a window under thirty years.

The precision was the point, and it came from the method rather than from the radiocarbon measurement alone. Wiggle-matching does not simply radiocarbon-date a sample; it dates a sequence of rings whose exact calendar spacing is known in advance, and lets the shape of that sequence’s carbon-14 variation lock onto the matching shape in the calibration curve, the way a barcode locks onto a reader. The technique can be far tighter than any single radiocarbon measurement, on one condition: that the rings being matched really do represent one calendar year each, correctly counted, none skipped and none counted twice. That condition belongs to dendrochronology, not to radiocarbon physics, and dendrochronologists do not extend it to every species. It holds for oak and for most conifers of temperate latitudes. It is known to fail in species that lay down more than one growth flush in a wet year, or none at all in a dry one. Olive sits on the known-to-fail side of that line (Cherubini et al. 2013).

Paolo Cherubini and eight co-authors tested the assumption rather than argued about it. They sent sections of Santorini olive wood, cut to include growth of the kind the 2006 dating had relied on, to five dendrochronology laboratories without telling them which rings the original study had assigned to which years. The labs disagreed with each other, and disagreed badly: ring counts for the same piece of wood varied by tens of rings, with annual boundaries that were sometimes faint, sometimes doubled within a single growing season, and sometimes simply absent where Mediterranean olive lays down no distinct line in a year too dry to mark one (Cherubini et al. 2013). None of this disputes the carbon-14 values Friedrich’s team measured. It disputes what the rings being measured were entitled to mean. A sequence that is not reliably one ring per year cannot anchor a wiggle-match to a precision the calibration curve does not also carry, whatever the carbon says underneath it.

This is not a mistake confined to one laboratory. It is the place where two fields, each managing the same inferential step under a different name, failed to compare notes. Dendrochronology has a decades-old, almost bureaucratic caution about species selection, built out of exactly this kind of blind cross-dating failure, and it treats whether a given wood can be counted as a question to settle before a date is reported, not after. Aegean Bronze Age chronology is a much older argument, run for most of a century on ceramic typology and Egyptian synchronisms, and it is used to taking a stated calendar date from an allied science and folding it in as a hard constraint once published, because that is how cross-dating has always worked on the historical side: a cartouche in the right tomb, a scarab in the right stratum, and a date is a date. When the two fields met over the olive branch, the archaeological side imported a number and treated the uncertainty already attached to it, ninety-five percent rather than certainty, as the only uncertainty on offer, because nothing in a century of ceramic cross-dating had trained anyone to ask whether a tree had been counted correctly first. The dendrochronological side published to the chronology debate’s audience without making the ring-count question the headline risk, because within dendrochronology species selection is so basic a step that flagging it reads as stating the obvious, which is exactly the kind of thing that goes missing crossing into a field where it is not obvious at all.

Sturt Manning and six co-authors, publishing in the same issue of Science, reached a similarly high date from a larger and more varied radiocarbon dataset, a hundred and twenty-seven measurements from sites across Crete, Rhodes, Turkey and Santorini, run through a Bayesian model rather than a single wiggle-match (Manning et al. 2006). That result does not depend on counting olive rings and is not undone by Cherubini’s finding. Its reception still ran into the same translation gap, approached from the other side. Malcolm Wiener, arguing for the lower, Egyptian-synchronised chronology, devoted a paper to the ways radiocarbon results are measured, calibrated, and then communicated to archaeologists in a form that reads as more settled than the statistics beneath it, a complaint less about any single date than about two expert communities handing a number across a disciplinary border and letting its meaning erode in transit (Wiener 2012). The complaint generalises past this one eruption: a radiocarbon result is a probability distribution built on calibration-curve assumptions that are themselves periodically revised, and a chronology built by treating it as a point value is building on a translation nobody checked.

The obvious response is that the system worked. Cherubini’s test exposed the olive branch’s weakness in 2013, and five years later Christopher Pearson and colleagues built an annual-resolution radiocarbon record for exactly this window from bristlecone pine and Irish oak, two species dendrochronology does trust to be counted correctly, and used it to show that the calibration curve itself carried a regional offset across these centuries, moving the best estimate for the eruption into the sixteenth century BC, closer to, though still earlier than, the archaeological range (Pearson et al. 2018). Science checked its own method and corrected it. On that reading the olive branch was never the real problem, only an early, badly anchored data point, and twelve years is not long for a one-off result to stand in a field this specialised before someone checks it.

That is true, and it does not reach the point. A result does not get treated as provisional while it is believed. For the twelve years between Friedrich’s paper and Pearson’s, 1627 to 1600 BC was the number in textbooks, encyclopaedia entries, and museum chronologies, cited as a precise, settled fact rather than as a figure resting on a ring count five laboratories could not agree on when asked to check it blind. Correction arriving eventually is not the same claim as the uncertainty having been visible at the time, and the twelve-year gap is exactly the cost of the caution that failed to travel: a problem dendrochronology already knew how to flag sat unflagged because it crossed into a field with no habit of asking the question. Nothing here says wiggle-matching is unreliable in general, or that the high Aegean chronology is wrong; Manning’s broader dataset may yet hold up regardless of what happened to one twig. The claim is narrower and more durable than either side of the dating dispute: when a method travels from the field that built its safeguards to a field that did not, the safeguard has to travel with the number, in the same paper, in a form the receiving field can actually use, or the next twelve years belong to whichever number sounds most finished.